DELTA: Delta Rule Learning Applet
- The Delta Rule Applet demonstrates a specific case of a single-layer network.
There are two inputs, x(1) and x(2), and three
connection weights, w(1) through w(3). The
value of x(3), the bias, is set to 1. The inputs and bias
are passed into a processing element that computes their
weighted sum (i.e. the summation of x(i)w(i)). Finally, the
output y is found based on the value of the sum; this
output is compared against the desired output T.
- Given a set of 3 initial connection weights and a learning rate, the applet
iterates using the Delta Rule algorithm until it finds a set of weights that
satisfies the AND or OR condition. You can also force the applet to stop
prematurely by specifying a maximum number of iterations.
- IMPORTANT:
The applet is written in Java 2, which is unfortunately not yet supported
in most popular browsers. You can circumvent this by installing Sun's
Java Runtime Environment.
This contains the company's Java plug-in and will enable you to view all Java 2
applets.
|
